Developing risk models for predicting incidence of diabetes and prediabetes in the first-degree relatives of Iranian patients with type 2 diabetes and comparison with the finnish diabetes risk score.

BACKGROUND

We aimed to develop risk models for predicting the onset of developing diabetes and prediabetes in the first-degree relatives (FDRs) of patients with type 2 diabetes, who have normal glucose tolerance (NGT).

MATERIALS AND METHODS

In this study, 1765 FDRs of patients with type 2 diabetes mellitus, who had NGT, were subjected to the statistical analysis. Diabetes risk factors, including anthropometric indices, physical activity, fast plasma glucose, plasma glucose concentrations 2-h after oral glucose administration, glycosylated hemoglobin (HbA1c), blood pressure, and lipid profile at the baseline were considered as independent variables. Kaplan-Meier, log-rank test, univariate, and multivariable proportional hazard Cox regression were used for the data analysis. The optimal cutoff value for risk score was created according to the receiver operating characteristic curve analysis.

RESULTS

The best diabetes predictability was achieved by a model in which waist-to-hip ratio, HbA1c, oral glucose tolerance test-area under the curve (OGTT-AUC), and the lipid profile were included. The best prediabetes risk model included HbA1c, OGTT-AUC, systolic blood pressure, and the lipid profile. The predictive ability of multivariable risk models was compared with fasting plasma glucose (FPG), HbA1c, and OGTT. The predictive ability of developed models was higher than FPG and HbA1c; however, it was comparable with OGTT-AUC alone. In addition, our study showed that the developed models predicted diabetes and OGTT-AUC better than the Finnish Diabetes Risk Score (FINDRISC).

CONCLUSION

We recommend regular monitoring of risk factors for the FDRs of patients with type 2 diabetes as an efficient approach for predicting and prevention of the occurrence of diabetes and prediabetes in future. Our developed diabetes risk score models showed precise prediction ability compared to the FINDRISC in Iranian population.